[c-nsp] iBGP v eBGP

Michael Robson Michael.Robson at manchester.ac.uk
Tue Apr 18 11:19:28 EDT 2006


> >The question: what are the benefits of using iBGP in the 
> core instead 
> >of just using eBGP everywhere? Someone asked me this and I couldn't 
> >answer the question to my own satisfaction let alone his.
> 
> eBGP is a BGP connection to a different AS than your 
> router's. Unless each router in your core is running a 
> different AS on it (ewwww), you can't run eBGP in the core. 
>
I missed out some information here. All the sites connected to 
the core only have this core through which to connect out to 
the Internet beyond and as such, each site will be using private
AS numbers which we can convert to our "real" AS number by using
a confederation or similar. Therefore, we will already have lots
of AS numbers "floating around" and so from this point of view, a
few more for each of the core routers, if they were to run eBGP
instead of iBGP wouldn't have been such of a big deal.

> I'm sure there are some convoluted ways someone could 
> invalidate what I said, but in general iBGP is what you get 
> when peering with routers in the same AS and what you want to do.
>
Not that convoluted using private AS numbers and stripping the AS
number off as it reaches the core or using a confederation as it
exits the core upstream.

> iBGP by default will only update other iBGP peers with routes 
> it is originating or has learned from an eBGP speaker. It 
> won't redistribute other learned iBGP routes which is why you 
> need a full mesh. 
>
This bit is what I was missing from my knowledge and what makes it
obvious to me now that iBGP is the sensible way forward.

Thanks,

Michael.
--
Michael Robson,           | Tel:  0161 275 6113
Networks,                 | Fax:  0161 275 6040
University of Manchester. | Email: Michael.Robson at manchester.ac.uk




More information about the cisco-nsp mailing list